pymatgen.analysis.structure_prediction package

Utilities to predict new structures.

Submodules

pymatgen.analysis.structure_prediction.dopant_predictor module

Predicting potential dopants.

get_dopants_from_shannon_radii(bonded_structure, num_dopants=5, match_oxi_sign=False)[source]

Get dopant suggestions based on Shannon radii differences.

Parameters:
  • bonded_structure (StructureGraph) – A pymatgen structure graph decorated with oxidation states. For example, generated using the CrystalNN.get_bonded_structure() method.

  • num_dopants (int) – The number of suggestions to return for n- and p-type dopants.

  • match_oxi_sign (bool) – Whether to force the dopant and original species to have the same sign of oxidation state. E.g. If the original site is in a negative charge state, then only negative dopants will be returned.

Returns:

Dopant suggestions, given as a dictionary with keys “n_type” and

”p_type”. The suggestions for each doping type are given as a list of dictionaries, each with they keys:

  • ”radii_diff”: The difference between the Shannon radii of the species.

  • ”dopant_species”: The dopant species.

  • ”original_species”: The substituted species.

Return type:

dict

get_dopants_from_substitution_probabilities(structure, num_dopants=5, threshold=0.001, match_oxi_sign=False) dict[source]

Get dopant suggestions based on substitution probabilities.

Parameters:
  • structure (Structure) – A pymatgen structure decorated with oxidation states.

  • num_dopants (int) – The number of suggestions to return for n- and p-type dopants.

  • threshold (float) – Probability threshold for substitutions.

  • match_oxi_sign (bool) – Whether to force the dopant and original species to have the same sign of oxidation state. E.g. If the original site is in a negative charge state, then only negative dopants will be returned.

Returns:

Dopant suggestions, given as a dictionary with keys “n_type” and

”p_type”. The suggestions for each doping type are given as a list of dictionaries, each with they keys:

  • ”probability”: The probability of substitution.

  • ”dopant_species”: The dopant species.

  • ”original_species”: The substituted species.

Return type:

dict

pymatgen.analysis.structure_prediction.substitution_probability module

This module provides classes for representing species substitution probabilities.

class SubstitutionPredictor(lambda_table=None, alpha=-5, threshold=0.001)[source]

Bases: object

Predicts likely substitutions either to or from a given composition or species list using the SubstitutionProbability.

Parameters:
  • lambda_table (dict) – Input lambda table.

  • alpha (float) – weight function for never observed substitutions

  • threshold (float) – Threshold to use to identify high probability structures.

composition_prediction(composition, to_this_composition=True)[source]

Get charged balanced substitutions from a starting or ending composition.

Parameters:
  • composition – starting or ending composition

  • to_this_composition – If true, substitutions with this as a final composition will be found. If false, substitutions with this as a starting composition will be found (these are slightly different)

Returns:

List of predictions in the form of dictionaries. If to_this_composition is true, the values of the dictionary will be from the list species. If false, the keys will be from that list.

list_prediction(species, to_this_composition=True)[source]
Parameters:
  • species – list of species

  • to_this_composition – If true, substitutions with this as a final composition will be found. If false, substitutions with this as a starting composition will be found (these are slightly different).

Returns:

List of predictions in the form of dictionaries. If to_this_composition is true, the values of the dictionary will be from the list species. If false, the keys will be from that list.

class SubstitutionProbability(**kwargs)[source]

Bases: object

This class finds substitution probabilities given lists of atoms to substitute. The inputs make more sense if you look through the from_defaults static method.

The substitution prediction algorithm is presented in: Hautier, G., Fischer, C., Ehrlacher, V., Jain, A., and Ceder, G. (2011) Data Mined Ionic Substitutions for the Discovery of New Compounds. Inorganic Chemistry, 50(2), 656-663. doi:10.1021/ic102031h

Parameters:
  • lambda_table – JSON table of the weight functions lambda if None, will use the default lambda.json table

  • alpha (float) – weight function for never observed substitutions.

as_dict()[source]

Get MSONable dict.

cond_prob(s1, s2)[source]

Conditional probability of substituting s1 for s2.

Parameters:
  • s1 – The variable specie

  • s2 – The fixed specie

Returns:

Conditional probability used by structure predictor.

cond_prob_list(l1, l2)[source]

Find the probabilities of 2 lists. These should include ALL species. This is the probability conditional on l2.

Parameters:
  • l1 – lists of species

  • l2 – lists of species

Returns:

The conditional probability (assuming these species are in l2)

classmethod from_dict(dct: dict) Self[source]
Parameters:

dct (dict) – Dict representation.

Returns:

Class

get_lambda(s1, s2)[source]
Parameters:
  • s1 (SpeciesLike) – Ion in 1st structure.

  • s2 (SpeciesLike) – Ion in 2nd structure.

Returns:

Lambda values

get_px(sp: SpeciesLike) float[source]
Parameters:

sp (SpeciesLike) – Species.

Returns:

Probability

Return type:

float

pair_corr(s1, s2)[source]

Pair correlation of two species.

Returns:

The pair correlation of 2 species

prob(s1, s2)[source]

Get the probability of 2 species substitution. Not used by the structure predictor.

Returns:

Probability of s1 and s2 substitution.

pymatgen.analysis.structure_prediction.substitutor module

This module provides classes for predicting new structures from existing ones.

class Substitutor(threshold=0.001, symprec: float = 0.1, **kwargs)[source]

Bases: MSONable

This object uses a data mined ionic substitution approach to propose compounds likely to be stable. It relies on an algorithm presented in Hautier, G., Fischer, C., Ehrlacher, V., Jain, A., and Ceder, G. (2011). Data Mined Ionic Substitutions for the Discovery of New Compounds. Inorganic Chemistry, 50(2), 656-663. doi:10.1021/ic102031h.

Use the substitution probability class to find good substitutions for a given chemistry or structure.

Parameters:
  • threshold – probability threshold for predictions

  • symprec – symmetry precision to determine if two structures are duplicates

  • kwargs – kwargs for the SubstitutionProbability object lambda_table, alpha

as_dict()[source]

Get MSONable dict.

charge_balanced_tol: float = 1e-09[source]
classmethod from_dict(dct: dict) Self[source]
Parameters:

dct (dict) – Dict representation.

Returns:

Class

get_allowed_species()[source]

Get the species in the domain of the probability function any other specie will not work.

pred_from_comp(composition) list[dict][source]

Similar to pred_from_list except this method returns a list after checking that compositions are charge balanced.

pred_from_list(species_list) list[dict][source]

There are an exceptionally large number of substitutions to look at (260^n), where n is the number of species in the list. We need a more efficient than brute force way of going through these possibilities. The brute force method would be:

output = [] for p in itertools.product(self._sp.species_list, repeat=len(species_list)):

if self._sp.conditional_probability_list(p, species_list) > self._threshold:

output.append(dict(zip(species_list, p)))

return output

Instead of that we do a branch and bound.

Parameters:

species_list – list of species in the starting structure

Returns:

list of dictionaries, each including a substitutions dictionary, and a probability value

pred_from_structures(target_species, structures, remove_duplicates=True, remove_existing=False) list[TransformedStructure][source]

Performs a structure prediction targeting compounds containing all of the target_species, based on a list of structure (those structures can for instance come from a database like the ICSD). It will return all the structures formed by ionic substitutions with a probability higher than the threshold.

Notes

If the default probability model is used, input structures must be oxidation state decorated. See AutoOxiStateDecorationTransformation

This method does not change the number of species in a structure. i.e if the number of target species is 3, only input structures containing 3 species will be considered.

Parameters:
  • target_species – a list of species with oxidation states e.g. [Species(‘Li+’), Species(‘Ni2+’), Species(‘O-2’)]

  • structures_list – list of dictionary of the form {‘structure’: Structure object, ‘id’: some id where it comes from} The id can for instance refer to an ICSD id.

  • remove_duplicates – if True, the duplicates in the predicted structures will be removed

  • remove_existing – if True, the predicted structures that already exist in the structures_list will be removed

Returns:

a list of TransformedStructure objects.

pymatgen.analysis.structure_prediction.volume_predictor module

Predict volumes of crystal structures.

class DLSVolumePredictor(cutoff=4.0, min_scaling=0.5, max_scaling=1.5)[source]

Bases: object

Data-mined lattice scaling (DLS) scheme that relies on data-mined bond lengths to predict the crystal volume of a given structure.

As of 2/12/19, we suggest this method be used in conjunction with min_scaling and max_scaling to prevent instances of very large, unphysical predicted volumes found in a small subset of structures.

Parameters:
  • cutoff (float) – cutoff radius added to site radius for finding site pairs. Necessary to increase only if your initial structure guess is extremely bad (atoms way too far apart). In all other instances, increasing cutoff gives same answer but takes more time.

  • min_scaling (float) – if not None, this will ensure that the new volume is at least this fraction of the original (preventing too-small volumes)

  • max_scaling (float) – if not None, this will ensure that the new volume is at most this fraction of the original (preventing too-large volumes).

get_predicted_structure(structure: Structure, icsd_vol=False)[source]

Given a structure, returns back the structure scaled to predicted volume.

Parameters:

structure (Structure) – structure w/unknown volume

Returns:

a Structure object with predicted volume

predict(structure: Structure, icsd_vol=False)[source]

Given a structure, returns the predicted volume.

Parameters:
  • structure (Structure) – a crystal structure with an unknown volume.

  • icsd_vol (bool) – True if the input structure’s volume comes from ICSD.

Returns:

a float value of the predicted volume.

class RLSVolumePredictor(check_isostructural=True, radii_type='ionic-atomic', use_bv=True)[source]

Bases: object

Reference lattice scaling (RLS) scheme that predicts the volume of a structure based on a known crystal structure.

Parameters:
  • check_isostructural – Whether to test that the two structures are isostructural. This algo works best for isostructural compounds. Defaults to True.

  • radii_type (str) – Types of radii to use. You can specify “ionic” (only uses ionic radii), “atomic” (only uses atomic radii) or “ionic-atomic” (uses either ionic or atomic radii, with a preference for ionic where possible).

  • use_bv (bool) – Whether to use BVAnalyzer to determine oxidation states if not present.

get_predicted_structure(structure: Structure, ref_structure)[source]

Given a structure, returns back the structure scaled to predicted volume.

Parameters:
  • structure (Structure) – structure w/unknown volume

  • ref_structure (Structure) – A reference structure with a similar structure but different species.

Returns:

a Structure object with predicted volume

predict(structure: Structure, ref_structure)[source]

Given a structure, returns the predicted volume.

Parameters:
  • structure (Structure) – structure w/unknown volume

  • ref_structure (Structure) – A reference structure with a similar structure but different species.

Returns:

a float value of the predicted volume